Ling H286 -- Term Project, Part 1, Perception data

Term Project, Part 1, Perception data

Ling H286, Autumn 2007, Ohio State University)

0. Due date

This part of the term project involves a perception done in class on Wednesday, October 10. If you did not do the perception experiment in class that day, do the experiment and e-mail your results table to the instructor by 5:00 on Friday, October 12, so that we can include them in file of all results that you and your group need to be analyzing for your group report.

Your group presentation will be in class on Wednesday, October 17, and your personal annotated copy of the presented material will due that day as well.

1. Do the experiment

The data for this first part of the term project are the results of running a perception experiment. To run the experiment, download the praat script Lx286perception.txt from the scripts directory to some folder on your hard drive. In that same folder, create a subfolder called "perceptionStimuli" and download the 40 sound files in the perceptionStimuli folder on our class web site to your perceptionStimuli folder. (Alternatively, download the termProjectPart1.zip file and unzip it somewhere.)

Read the Lx286perception.txt file into Praat. It will open as an object of type ExperimentMFC in your "Objects:" window. Plug in your headphones and make sure you have audio, and then click on the "Run" command button to run the experiment.

A large window will fill your screen, with text in telling you to "Click to start." Immediately after you click, you will hear a token of the word hod or of the word hawed. Your task is decide which you heard. If you have to guess, make the best guess that you can, but don't worry if it seems impossible. Different people have different pronunciations of these words and some of those pronunciations may not even sound like either word to you. Also, as some of you said in class when we made the recordings, not all of you make a difference between these two words. So, just pick the one it sounds closer to, click on that button and then click on the "okay" button. (You also have the option of clicking the "replay" button to play the recording just once more before you decide.)

When you have finished listening to all 40 tokens, close the experiment window. There should now be a ResultsMFC object with the same name as the ExperimentMFC object in your Praat "Objects:" window. Highlight that ResultsMFC object and rename it with your name. For example, if you are Alexa, click on the "Rename..." button just below the objects list and type "Alexa" in the "New name:" box in the GUI that comes up.

Then click on the "Collect to Table" command button. This will create a Table object named allResults. Use the "Rename..." command button again to rename this object and then use the "Write to table file..." command under the "Write" pull-down menu at the top of the window to write your results to a Table file called by your name. For example, Alexa's file will be called Alexa.txt. Give this file to Mary, to collate together with the other results for the class into one large results file.

2. Analyze the results

The almost final collated results file is Au2007perception.txt in the termProjectPart1Results directory of our course web page. It has a header row and currently has 720 rows of data, one row for each of the 40 response that each of you made in this experiment on Wednesday. (There will 40 more rows when Chris has done the experiment.) There are six categorical variables in each row for:

listenerDistinguishes -- "yes" or "no" for whether the perosn who made the response in that row reports distinguishing hod and hawed
listener -- the name of the person who made the response in that row
talkerDistinguishes -- like the first column but for the talker
talker -- the name of the person who produced the stimulus in that row
stimulus -- the word that the talker was saying
response -- the word that the listener chose to click on in the experiment

Work with your group to analyze these results. There is R code in the script perceptionResults.R in the scripts directory that might be useful to you. Work out a strategy for using the perception results to answer the following question:

Question: When people do not make a distinction in their own productions can they (acting as listeners in an experiment like this) still identify the words produced people who do make the distinction?

Your strategy should make predictions about accuracy of identification for different subsets of the stimuli and different subsets of listeners, as identified in the following table.

		talker who produced stimulus ...
		makes distinction	does not make
listener who responded ...	has distinction	set 1	set 2
listener who responded ...	does not have	set 3	set 4

You will have a chance to work together as a group in class on some of this on the day that we do the in-class experiment and also on Monday, October 15, when you can ask the instructor questions.

3. Make the oral report

Decide which two members of your will make the oral presentation of your group report, remembering that each of you should present at least once. Aim for a presentation of 5-7 minutes and no more than five slides, including all of the data slides.